This paper reports a surprising reversal in the effect of advertising repetition over time. A field study shows higher annoyance with a more frequently advertised brand at the time of advertising, but greater preference for this same brand several weeks later. A longitudinal online experiment replicates the reversal in brand preference across four time points and tests an underlying mechanism for this reversal. It shows that initial annoyance with frequent ad repetition is highly susceptible to decay over time, whereas brand memory remains relatively stable. Through these two processes, brands with heavier advertising exposure move over time from lower preference to higher preference. Finally, a third experiment demonstrates a crucial condition for this reversal — product relevance at the time of consideration. This study shows that the reversal occurs only if at the later time the product category is relevant. We discuss the substantial implications of these findings for marketing theory and practice.